Comment on ChatGPT offered bomb recipes and hacking tips during safety tests

einkorn@feddit.org ⁨2⁩ ⁨weeks⁩ ago

• BussyGyatt@feddit.org ⁨2⁩ ⁨weeks⁩ ago

  well, yes, but the point is they specifically asked chatgpt not to produce bomb manuals when they were training it. or thought they did; evidently that’s not what they actually did.

  source

  • otter@lemmy.ca ⁨2⁩ ⁨weeks⁩ ago

    specifically trained chatgpt not

    Often this just means appending “do not say X” to the start of every message, which then breaks down when the user says something unexpected right afterwards

    I think moving forward

    • companies selling generative AU need to be more honest about the capabilities of the tool
    • people need to understand that it’s a very good text prediction engine being used for other tasks

    source

    • panda_abyss@lemmy.ca ⁨2⁩ ⁨weeks⁩ ago

      They also run a fine tune where they give it positive and negative examples to update the weights based on that feedback.

      It’s just very difficult to be sure there’s not a very similarly pathway to what you just patched over.

      source

      • spankmonkey@lemmy.world ⁨2⁩ ⁨weeks⁩ ago

        It isn’t very difficult, it is fucking impossible. There are far too many permutations to be manually countered.

        source

        • -> View More Comments
    • BussyGyatt@feddit.org ⁨2⁩ ⁨weeks⁩ ago

      my original comment before editing read something like “they specifically asked chatgpt not to produce bomb manuals when they trained it” but i didn’t want people to think I was anthropomorphizing the llm.

      source